UNIXLONG at SemEval-2020 Task 6: A Joint Model for Definition Extraction

Proceedings of the 14th International Workshop on Semantic Evaluation, pages 730–736 Barcelona, Spain (Online), December 12, 2020. 730 UNIXLONG at SemEval-2020 Task 6: A Joint Model for Definition Extraction Shu-Yi Xie , Jian Ma , Hai-Qin Yang, Lian-Xin Jiang , Yang Mo , Jian-Ping Shen Ping An Life Insurance, Lt AI Department ShenZhen, China {XIESHUYI542,MAJIAN446,YANGHAIQIN260,JIANGLIANXIN769, MOYANG853,SHENJIANPING324}@pingan.com.cn Abstract Definition Extraction is the task to automatically extract terms and their definitions from text. In recent years, it attracts wide interest from NLP researchers. This paper describes the unixlong team’s system for the SemEval 2020 task6: DeftEval: Extracting term-definition pairs in free text. The goal of this task is to extract definition, word level BIO tags and relations. This task is challenging due to the free style of the text, especially the definitions of the terms range across several sentences and lack explicit verb phrases. We propose a joint model to train the tasks of definition extraction and the word level BIO tagging simultaneously. We design a creative format input of BERT to capture the location information between entity and its definition. Then we adjust the result of BERT with some rules. Finally, we apply TAG ID, ROOT ID, BIO tag to predict the relation and achieve macro-averaged F1 score 1.0 which rank first on the official test set in the relation extraction subtask. 1 Introduction Definition extraction is a hot research topic in natural language processing (NLP) that aims to identify terms and their corresponding definitions in unstructured text. However, it is a challenging problem to extract term-definition pairs from free and semi-structured text, especially those whose term-definition pair span crosses a sentence boundary and those lacking explicit definition verb phrases. SemEval is a yearly organized NLP competition. The SemEval 2020 - Task 6, “DeftEval: Extracting term-definition pairs in free text”, aims to extract definition from the DEFT (Sasha et al., 2019) corpus, which contains roughly 7,000 sets of 3-sentence groupings extracted from textbooks of various topics from cnx.org. Each sentence set reflects a context window around which the author of the original text marked a bolded word to indicate a key term. DeftEval consists of three tasks: 1) Subtask 1: Sentence Classification. Given a sentence, it is to classify whether the sentence contains a definition or not. This is a traditional definition extraction task. 2) Subtask 2: Sequence Labeling. It is to label each token with the BIO tag according to the corpus’ tag specification. 3) Subtask 3: Relation Classification. Given the tagging results, it is to label the relations between each tag according to the corpus’ relation specification. This paper describes our system and participation in all subtasks of SemEval 2020 task6. The challenges of DeftEval include: 1) The DEFT corpus contains various topics, such as biology, history, physics, psychology, economics, sociology, and government, etc. It is challenging and requires deep background knowledge to understand the professional definition from these topics. 2) The sentences in the DEFT corpus are extremely complicated. A term and its definition may come from several …

Content truncated to 500 words